Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Research idea? re: weight optimization

Author: Antonio Dieguez

Date: 21:23:01 12/29/01

Go up one level in this thread


On December 29, 2001 at 21:57:44, Tom Kerrigan wrote:

>On December 29, 2001 at 20:42:18, Martin Giepmans wrote:
>
>>On December 29, 2001 at 17:14:45, Tom Kerrigan wrote:
>>
>>>On December 29, 2001 at 09:37:55, Martin Giepmans wrote:
>>>
>>>>On December 29, 2001 at 05:11:26, Tom Kerrigan wrote:
>>>>
>>>>>Here's my idea.
>>>>>
>>>>>You have a position and you want your program to play a certain move (which it
>>>>>presumably isn't playing). You run this algorithm:
>>>>>
>>>>>1. Search the position, get a PV. The evaluation of the last position of the PV
>>>>>is eval(1).
>>>>>2. Search only the move that you want your program to make, get a PV. This
>>>>>end-point evaluation is eval(2).
>>>>>3. Figure out which eval terms are different between eval(1) and eval(2).
>>>>>Decrease the weights of all the different eval(1) terms slightly. Increase the
>>>>>eval(2) terms slightly.
>>>>>4. Repeat until the program plays the move you want.
>>>>>
>>>>>You could run this on lots of positions from GM games, to get your program to
>>>>>play like a GM. (At least in some positions, heh.)
>>>>>
>>>>>Has this been done before? Are there any glaring problems with this idea? Does
>>>>>anybody want to try this? If so, I'd like some credit for it. If not, I'll
>>>>>probably get around to trying it sometime...
>>>>>
>>>>>-Tom
>>>>
>>>>
>>>>I think one of the bigger problems with this idea is that the minimax-eval
>>>>at the root does not only depend on the eval of the last position of the PV.
>>>>It usually also depends on the eval of many other positions in the search-tree.
>>>
>>>Hence iterating the searches.
>>>
>>>-Tom
>>
>>
>>Hi Tom,
>>
>>I'm afraid iterating doesn't solve this, because many positions
>>that influence the search (and ultimately the move that is chosen
>>at the root) will never appear at the end of a PV.
>>So I guess that looking at endpoints is simply not enough.
>
>I'm not sure what you mean. The only "influence" that other positions have on
>the root node is that they're not good enough to end up at the end of the PV...

may be they should be good enough...



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.